The Epistemological Roots of AI Hallucination and Its Virtue-Based Governance

Qing Huang

doi:10.6918/IJOSSER.202606_9(6).0003

Authors

Qing Huang

DOI:

https://doi.org/10.6918/IJOSSER.202606_9(6).0003

Keywords:

AI hallucination; large language models; embodied cognition; virtue epistemology; cognitive responsibility.

Abstract

Large language models (LLMs) hallucinate, and they do so pervasively. This paper treats that fact as more than a technical defect. Drawing on phenomenology, epistemology, and the philosophy of science and technology, it argues that hallucination is an epistemological problem—what surfaces when generative AI operates without any relation to the world or any intentional structure. I begin with the generative mechanism itself: probabilistic next-token prediction, and the tension it creates between statistical correlation among symbols and semantic truth. A comparison of human and machine cognition then locates the deeper limit of the machine side in its lack of intentionality, embodiment, and practical feedback. Against this background I assess four governance pathways—Retrieval-Augmented Generation (RAG), reinforcement-learning alignment, embodied intelligence, and neuro-symbolic AI—and the point at which each stalls. I further argue that current alignment techniques risk substituting preference for truth, and that sycophantic alignment can breed intellectual sloth and epistemic arrogance in users. On this basis the paper reconstructs the governance of hallucination within virtue epistemology, drawing individual prudence, institutional empowerment, and system design into one shared mechanism of human–machine epistemic responsibility—a way of protecting human cognitive sovereignty in the age of algorithms.

Downloads

Download data is not yet available.

References

[1] UBS Chief Investment Office GWM. (2023). Information technology: Let's chat about ChatGPT. UBS.

[2] Royal Swedish Academy of Sciences. (2024). Scientific background: Computational protein design and protein structure prediction. Royal Swedish Academy of Sciences.

[3] Huang, L., Yu, W., Ma, W., et al. (2025). A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Transactions on Information Systems, 43(2), 42:1–42:55.

[4] Vaswani, A., Shazeer, N., Parmar, N., et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.

[5] Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March 3–10). On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (Virtual Event, Canada, pp. 610–623).

[6] Hicks, M. T., Humphries, J., & Slater, J. (2024). ChatGPT is bullshit. Ethics and Information Technology, 26(2), 38.

[7] Husserl, E. (2017). Logical investigations: Volume II: Investigations in phenomenology and the theory of knowledge, Part I. Commercial Press. pp. 792–793, 824. (In Chinese)

[8] Merleau-Ponty, M. (2001). Phenomenology of perception. Commercial Press. pp. 126–144. (In Chinese)

[9] Heidegger, M. (2018). Being and time (2nd rev. Chinese ed.). Commercial Press. pp. 88–94. (In Chinese)

[10] Heidegger, M. (2015). On the way to language. Commercial Press. pp. 10, 12. (In Chinese)

[11] Mandelkern, M., & Linzen, T. (2024). Do language models' words refer? Computational Linguistics, 50(3), 1191–1200.

[12] Kalai, A. T., Nachum, O., Vempala, S. S., & Zhang, E. (2025). Why language models hallucinate. arXiv:2509.04664.

[13] Dong, C. (2023). The nature and limits of artificial intelligence from the perspective of machine cognitive opacity. Social Sciences in China, (5), 44–66.

[14] Burrell, J. (2016). How the machine 'thinks': Understanding opacity in machine learning algorithms. Big Data & Society, 3(1), 1–12.

[15] Floridi, L. (2023). AI as agency without intelligence: On ChatGPT, large language models, and other generative models. Philosophy & Technology, 36(1), 15.

[16] Lewis, P., Perez, E., Piktus, A., et al. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459–9474.

[17] Asai, A., Wu, Z., Wang, Y., Sil, A., & Hajishirzi, H. (2024). Self-RAG: Learning to retrieve, generate, and critique through self-reflection. In Proceedings of the International Conference on Learning Representations (Vienna, Austria).

[18] Ouyang, L., Wu, J., Jiang, X., et al. (2022). Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35, 27730–27744.

[19] Bai, Y., Jones, A., Ndousse, K., et al. (2022). Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv:2204.05862.

[20] Rafailov, R., Sharma, A., Mitchell, E., Manning, C. D., Ermon, S., & Finn, C. (2023). Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36, 53728–53741.

[21] Sharma, M., Tong, M., Korbak, T., et al. (2023). Towards understanding sycophancy in language models. arXiv:2310.13548.

[22] LeCun, Y. (2022). A path towards autonomous machine intelligence. OpenReview.

[23] Ha, D., & Schmidhuber, J. (2018). World models. arXiv:1803.10122.

[24] Dreyfus, H. L. (2007). Why Heideggerian AI failed and how fixing it would require making it more Heideggerian. Philosophical Psychology, 20(2), 247–268.

[25] Barsalou, L. W. (2008). Grounded cognition. Annual Review of Psychology, 59, 617–645.

[26] Driess, D., Xia, F., Sajjadi, M. S. M., et al. (2023, July 23–29). PaLM-E: An embodied multimodal language model. In Proceedings of the 40th International Conference on Machine Learning (Honolulu, USA, Vol. 202, pp. 8469–8488). PMLR.

[27] d'Avila Garcez, A., & Lamb, L. C. (2023). Neurosymbolic AI: The 3rd wave. Artificial Intelligence Review, 56(11), 12387–12406.

[28] Zagzebski, L. T. (1996). Virtues of the mind: An inquiry into the nature of virtue and the ethical foundations of knowledge. Cambridge University Press. pp. 114–115, 271.

[29] Sun, W. (2017). Value reflections on artificial intelligence. Philosophical Research, (10), 120–126.

[30] Liu, Y., & Wang, C. (2023). Human-machine relations in the age of intelligence: Toward a selectionist theory of technological control. Global Media Journal, 10(3), 5–21.

[31] Pang, Z., Xue, L., & Liang, Z. (2022). Artificial intelligence governance: Cognitive logic and paradigm transcendence. Science of Science and Management of Science and Technology, 43(9), 3–18.

[32] High-Level Expert Group on Artificial Intelligence. (2019). Ethics guidelines for trustworthy AI. European Commission.

[33] Mora-Cantallops, M., Sanchez-Alonso, S., Garcia-Barriocanal, E., & Sicilia, M. A. (2021). Traceability for trustworthy AI: A review of models and tools. Big Data and Cognitive Computing, 5(2), 20.